lkml.org 
[lkml]   [2021]   [Jul]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
Subject[PATCH 2/2] kvm: mmu/x86: Add detailed page size stats
From
Existing KVM code tracks the number of large pages regardless of their
sizes. Therefore, when large page of 1GB (or larger) is adopted, the
information becomes less useful because lpages counts a mix of 1G and 2M
pages.

So bridge the gap and provide a comprehensive page stats of all sizes from
4K to 512G.

Suggested-by: Ben Gardon <bgardon@google.com>
Suggested-by: Jing Zhang <jingzhangos@google.com>
Signed-off-by: Mingwei Zhang <mizhang@google.com>
---
arch/x86/include/asm/kvm_host.h | 11 ++++++++-
arch/x86/kvm/mmu.h | 2 ++
arch/x86/kvm/mmu/mmu.c | 43 +++++++++++++++++++++++----------
arch/x86/kvm/mmu/tdp_mmu.c | 10 +++-----
arch/x86/kvm/x86.c | 6 ++++-
5 files changed, 51 insertions(+), 21 deletions(-)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 974cbfb1eefe..1b7b024f9573 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1206,9 +1206,18 @@ struct kvm_vm_stat {
u64 mmu_recycled;
u64 mmu_cache_miss;
u64 mmu_unsync;
- u64 lpages;
+ atomic64_t lpages;
u64 nx_lpage_splits;
u64 max_mmu_page_hash_collisions;
+ union {
+ struct {
+ atomic64_t pages_4k;
+ atomic64_t pages_2m;
+ atomic64_t pages_1g;
+ atomic64_t pages_512g;
+ };
+ atomic64_t pages[4];
+ } page_stats;
};

struct kvm_vcpu_stat {
diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
index 83e6c6965f1e..56d9c947a0cd 100644
--- a/arch/x86/kvm/mmu.h
+++ b/arch/x86/kvm/mmu.h
@@ -240,4 +240,6 @@ static inline bool kvm_memslots_have_rmaps(struct kvm *kvm)
return smp_load_acquire(&kvm->arch.memslots_have_rmaps);
}

+void kvm_update_page_stats(struct kvm *kvm, u64 spte, int level, int delta);
+
#endif
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index c45ddd2c964f..9ba25f00ca2b 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -588,16 +588,33 @@ static bool mmu_spte_update(u64 *sptep, u64 new_spte)
return flush;
}

+void kvm_update_page_stats(struct kvm *kvm, u64 spte, int level, int count)
+{
+ if (!is_last_spte(spte, level))
+ return;
+ /*
+ * If the backing page is a large page, update the lpages stat first,
+ * then log the specific type of backing page. Only log pages at highter
+ * levels if they are marked as large pages. (As opposed to simply
+ * pointing to another level of page tables.).
+ */
+ if (is_large_pte(spte))
+ atomic64_add(count, (atomic64_t *)&kvm->stat.lpages);
+ atomic64_add(count,
+ (atomic64_t *)&kvm->stat.page_stats.pages[level-1]);
+}
+
/*
* Rules for using mmu_spte_clear_track_bits:
* It sets the sptep from present to nonpresent, and track the
* state bits, it is used to clear the last level sptep.
* Returns non-zero if the PTE was previously valid.
*/
-static int mmu_spte_clear_track_bits(u64 *sptep)
+static int mmu_spte_clear_track_bits(struct kvm *kvm, u64 *sptep)
{
kvm_pfn_t pfn;
u64 old_spte = *sptep;
+ int level = sptep_to_sp(sptep)->role.level;

if (!spte_has_volatile_bits(old_spte))
__update_clear_spte_fast(sptep, 0ull);
@@ -607,6 +624,8 @@ static int mmu_spte_clear_track_bits(u64 *sptep)
if (!is_shadow_present_pte(old_spte))
return 0;

+ kvm_update_page_stats(kvm, old_spte, level, -1);
+
pfn = spte_to_pfn(old_spte);

/*
@@ -984,9 +1003,10 @@ static void __pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
}
}

-static void pte_list_remove(struct kvm_rmap_head *rmap_head, u64 *sptep)
+static void pte_list_remove(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
+ u64 *sptep)
{
- mmu_spte_clear_track_bits(sptep);
+ mmu_spte_clear_track_bits(kvm, sptep);
__pte_list_remove(sptep, rmap_head);
}

@@ -1119,7 +1139,7 @@ static u64 *rmap_get_next(struct rmap_iterator *iter)

static void drop_spte(struct kvm *kvm, u64 *sptep)
{
- if (mmu_spte_clear_track_bits(sptep))
+ if (mmu_spte_clear_track_bits(kvm, sptep))
rmap_remove(kvm, sptep);
}

@@ -1129,7 +1149,6 @@ static bool __drop_large_spte(struct kvm *kvm, u64 *sptep)
if (is_large_pte(*sptep)) {
WARN_ON(sptep_to_sp(sptep)->role.level == PG_LEVEL_4K);
drop_spte(kvm, sptep);
- --kvm->stat.lpages;
return true;
}

@@ -1386,7 +1405,7 @@ static bool kvm_zap_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
while ((sptep = rmap_get_first(rmap_head, &iter))) {
rmap_printk("spte %p %llx.\n", sptep, *sptep);

- pte_list_remove(rmap_head, sptep);
+ pte_list_remove(kvm, rmap_head, sptep);
flush = true;
}

@@ -1421,13 +1440,13 @@ static bool kvm_set_pte_rmapp(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
need_flush = 1;

if (pte_write(pte)) {
- pte_list_remove(rmap_head, sptep);
+ pte_list_remove(kvm, rmap_head, sptep);
goto restart;
} else {
new_spte = kvm_mmu_changed_pte_notifier_make_spte(
*sptep, new_pfn);

- mmu_spte_clear_track_bits(sptep);
+ mmu_spte_clear_track_bits(kvm, sptep);
mmu_spte_set(sptep, new_spte);
}
}
@@ -2232,8 +2251,6 @@ static int mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
if (is_shadow_present_pte(pte)) {
if (is_last_spte(pte, sp->role.level)) {
drop_spte(kvm, spte);
- if (is_large_pte(pte))
- --kvm->stat.lpages;
} else {
child = to_shadow_page(pte & PT64_BASE_ADDR_MASK);
drop_parent_pte(child, spte);
@@ -2690,10 +2707,10 @@ static int mmu_set_spte(struct kvm_vcpu *vcpu, u64 *sptep,

pgprintk("%s: setting spte %llx\n", __func__, *sptep);
trace_kvm_mmu_set_spte(level, gfn, sptep);
- if (!was_rmapped && is_large_pte(*sptep))
- ++vcpu->kvm->stat.lpages;

if (!was_rmapped) {
+ kvm_update_page_stats(vcpu->kvm, *sptep,
+ sptep_to_sp(sptep)->role.level, 1);
rmap_count = rmap_add(vcpu, sptep, gfn);
if (rmap_count > RMAP_RECYCLE_THRESHOLD)
rmap_recycle(vcpu, sptep, gfn);
@@ -5669,7 +5686,7 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
if (sp->role.direct && !kvm_is_reserved_pfn(pfn) &&
sp->role.level < kvm_mmu_max_mapping_level(kvm, slot, sp->gfn,
pfn, PG_LEVEL_NUM)) {
- pte_list_remove(rmap_head, sptep);
+ pte_list_remove(kvm, rmap_head, sptep);

if (kvm_available_flush_tlb_with_range())
kvm_flush_remote_tlbs_with_address(kvm, sp->gfn,
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index caac4ddb46df..24bd7f03248c 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -446,12 +446,10 @@ static void __handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,

trace_kvm_tdp_mmu_spte_changed(as_id, gfn, level, old_spte, new_spte);

- if (is_large_pte(old_spte) != is_large_pte(new_spte)) {
- if (is_large_pte(old_spte))
- atomic64_sub(1, (atomic64_t*)&kvm->stat.lpages);
- else
- atomic64_add(1, (atomic64_t*)&kvm->stat.lpages);
- }
+ if (is_large_pte(old_spte) && !is_large_pte(new_spte))
+ kvm_update_page_stats(kvm, old_spte, level, -1);
+ else if (!is_large_pte(old_spte) && is_large_pte(new_spte))
+ kvm_update_page_stats(kvm, new_spte, level, 1);

/*
* The only times a SPTE should be changed from a non-present to
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 8166ad113fb2..23444257fcbd 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -237,7 +237,11 @@ const struct _kvm_stats_desc kvm_vm_stats_desc[] = {
STATS_DESC_ICOUNTER(VM, mmu_unsync),
STATS_DESC_ICOUNTER(VM, lpages),
STATS_DESC_ICOUNTER(VM, nx_lpage_splits),
- STATS_DESC_PCOUNTER(VM, max_mmu_page_hash_collisions)
+ STATS_DESC_PCOUNTER(VM, max_mmu_page_hash_collisions),
+ STATS_DESC_ICOUNTER(VM, page_stats.pages_4k),
+ STATS_DESC_ICOUNTER(VM, page_stats.pages_2m),
+ STATS_DESC_ICOUNTER(VM, page_stats.pages_1g),
+ STATS_DESC_ICOUNTER(VM, page_stats.pages_512g)
};
static_assert(ARRAY_SIZE(kvm_vm_stats_desc) ==
sizeof(struct kvm_vm_stat) / sizeof(u64));
--
2.32.0.402.g57bb445576-goog
\
 
 \ /
  Last update: 2021-07-21 07:15    [W:0.059 / U:0.056 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site